← Back to team overview

dolfin team mailing list archive

Re: Image to Function data structure conversion

 

On Fri, Jan 30, 2009 at 03:34:20AM +0000, A Navaei wrote:
> > I'd recommend looking at ImageMagick for reading image data. It has a
> > standard C++ library for reading/manipulating image data. ITK seems a
> > bit overkill for this.
> >
> > It's available in Ubuntu in the package libmagick++9-dev. One could
> > even consider adding a helper class to DOLFIN that allows simple
> > creation of Functions from image files:
> >
> >  f = Function("image.png")
> 
> ImageMagick is not really aimed for scientific applications, eg, it
> doesn't seem if it supports any dimensions other than 2. It would be
> fine if all we wanted was image data I/O, in that case even python
> image library (PIL) would work.

Good point, I wasn't thinking about 3D.

> In a real-world application, we will need more than image data I/O and
> it is desirable to perform some pre/post-processing of image data
> before/after data flow to dolfin. Since many of these processing
> algorithms are already implemented in itk, perhaps we shouldn't
> re-invent the wheel in less advanced libraries. Moreover, there would
> be an interest from the itk community to use dolfin as the
> finite-element engine since the current implementation of FEM in itk
> is known to be slow. As Luca mentioned earlier in this thread, this
> connection can be a starting point for many other developments.

I wasn't aware of this opportunity. I look forward to seeing where
this goes.

> I am in the process of setting up a build system in opensuse build
> service (https://build.opensuse.org) for a python interface to itk
> (and itk-dolfin connector). The build bot will produce packages for
> many linux platforms, including debian/ubuntu, and architectures at
> once (any plans for fenics to work on this build service, or are you
> happy with your buildbot?).

I'm pretty happy with what we have now, although we're not bulding
RPMs. Have you looked at Dorsal? It builds all of FEniCS on a
multitude of platforms?

  http://www.fenics.org/wiki/Dorsal

-- 
Anders

Attachment: signature.asc
Description: Digital signature


References